Enhancing Robotic Collaborative Tasks Through Contextual Human Motion Prediction and Intention Inference

Area of Science:

• Robotics and Computer Vision
• Artificial Intelligence
• Human-Robot Interaction

Background:

• Current 3D human motion prediction models use datasets that don't reflect human movement near robots.
• This data distribution gap limits the effectiveness of robots in real-world collaborative tasks.
• Incorporating task context and human willingness to collaborate can improve prediction accuracy and robot navigation.

Purpose of the Study:

• To propose a deep learning architecture for predicting both 3D human body motion and human intention in collaborative tasks.
• To enhance the capabilities of robots in navigating and succeeding in collaborative scenarios.
• To develop a flexible model adaptable to various tasks with different input requirements.

Main Methods:

• A deep learning architecture utilizing a multi-head attention mechanism was developed.
• The model takes human motion and task context as inputs to predict human body motion and infer human intention.
• The architecture was validated on collaborative object handover and grape harvesting tasks, with task-specific input variations.

Main Results:

• The proposed architecture successfully predicts 3D human motion and infers human intention for collaborative tasks.
• User studies for the handover task showed improved human perception of the robot's sociability, naturalness, security, and comfort when using the prediction.
• The model demonstrated promising results in the collaborative grape harvesting task, showcasing its potential for real-world applications.

Conclusions:

• The developed deep learning architecture effectively integrates human motion prediction and intention inference for enhanced human-robot collaboration.
• The model's flexibility in handling diverse inputs makes it adaptable for various real-world collaborative tasks.
• This research contributes to more intuitive, safe, and efficient human-robot interaction in shared environments.